Is postintubation hypotension really a determinant of increased mortality in traumatic patients?

    loading  Checking for direct PDF access through Ovid


To the Editor:
The recent article by Green et al1 determining the prevalence of postintubation hypotension (PIH) in trauma patients and the association of PIH with patient outcomes was of great interest. By multivariate logistic regressions analysis, they showed that PIH was associated with increased emergency department mortality and in-hospital mortality. However, this is a retrospective study, which potentially introduces a number of confounders. Other than the limitations described in the discussion section, we note other issues of this study making interpretation of their results questionable.
First, the authors only observed the incidence of PIH, but not the duration of PIH. The available evidence shows that the duration of hypotension is significantly associated with the occurrence of vital organ injury.2 Moreover, it was unclear whether any patient underwent surgical treatment and whether any major complication occurred during in-hospital period. Consequently, it is difficult to estimate the extent to which these factors might have influenced the occurrence of in-hospital death. In fact, both surgical burden (such as operative time and extend, mass blood loss and transfusion) and major complications (such as cardiac failure, pulmonary embolism, pneumonia, deep venous thrombosis and acute renal failure) are important determinants for in-hospital mortality of traumatic patients.3
Second, in a retrospective study, the multivariable logistic regression analysis is useful for adjusting the relative confounders and controlling selection biases. To obtain the true inferences of multivariable logistic regression analysis for adjusted odds ratios of measured outcomes, however, all known risk factors affecting measured outcome must be taken into account within the model. If an important risk factor is missed, the multivariate adjustment for odds ratios of measured outcomes can be biased and a spurious association between intervention and outcome of interest may be obtained. When using multivariate logistic regression analysis to determine the association between PIH and patient outcomes in this study, the model only included the patient characteristics (age, sex, injury type, injury severity, traumatic brain injury, volume of fluids given, medications administered) and provider characteristics, but not patient comorbidities. It has been shown that comorbidity score is a good predictor for in-hospital mortality of traumatic patients, and inclusion of comorbidity index in the model based on the patient characteristics can improve outcome prediction of traumatic patients. Furthermore, there is no interaction between age and comorbidity, suggesting that both independently increase vulnerability to mortality after injury.4,5 Thus, we argue that no inclusion of comorbidities in multivariable logistic regression models would have biased the true effect of PIH on short-term mortality of patients in this study.
Third, patients with PIH were older and more likely to have an Injury Severity Score of 12 or greater than those without PIH. This suggests that patients with PIH were likely to have worse health status and heavier trauma burden. In our opinion, no matter how refined the adjustment is for differences in patient characteristics and trauma severity, it is never possible to ensure a complete adjustment for differences among traumatic patients with and without PIH. PIH may be only a synthetical manifestation of worse health status and heavier trauma burden that can markedly increase mortality of traumatic patients. Moreover, the multivariable logistic regression analysis is probably inadequate to differentiate whether PIH is a true determinant of mortality or simply synthetical manifestation of worse health status and heavier trauma burden that can significantly increase mortality. Thus, we agree with the authors that statistical association between PIH and short-term mortality in this study cannot be used to imply causality.
    loading  Loading Related Articles